1. Field of the Invention
The present invention relates to an optical inspection system for performing optical inspection of a wafer and a semiconductor substrate, and more particularly, to an optical inspection system for detecting defects on a wafer, which are generated in the manufacturing process of semiconductor integrated circuits.
2. Description of the Related Art
A method of comparing an inspection target image data with a reference image data to check the integrity of an inspection target based on any accidental difference between them is conventionally known as optical inspection method for detecting conformity of pattern formed on an inspection target (a semiconductor wafer) and defects such as foreign substances and scratches on the pattern.
For example, through image data input means for gray scale image data representing an appearance of the inspection target in gray scale, and statistical processing means for calculating, as statistical data, reference image data and dispersion data based on a plurality of gray scale image data representing appearances of a plurality of inspection targets in gray scale, the statistical data calculated by the statistical processing means, discriminant data for determining acceptability of the inspection target, which is set in advance, and the gray scale image data of the inspection target, which is an input from the image data input means, are used to check integrity of the inspection target for each pixel, thereby determining acceptability of the inspection target.
As a method for determining acceptability, there is a method of calculating dispersion data and an average value of reference image data based on gray scale image data of a plurality of acceptable inspection targets in the statistical processing means, and calculating the range of acceptability based on the dispersion data and the average value of the reference image data, thereby determining acceptability of the inspection target. Further, the range of acceptability is set to different values for each of a plurality of areas with different position coordinates of gray scale image data, thereby determining acceptability of the inspection target with different inspection accuracy for each of the plurality of areas with different position coordinates of the gray scale image data (see, for example, JP 2004-93338 A).
In the conventional art described above, the standard for acceptability can be set in advance with different accuracy for each of the plurality of areas with different position coordinates. However a semiconductor device is, in general, manufactured by forming various thin films such as a silicon oxide film, a silicon nitride film, and polysilicon film on a semiconductor wafer, thickness variation in each of the various thin films occurs between lots, in a lot, and even in a wafer depending on the capacity of production equipment, which leads to a problem that a gray scale image of each chip on a semiconductor wafer changes, and in the case where film thickness of various thin films deviates from the range of the film thickness variation estimated in advance, the chip is detected as a defective chip despite absence of foreign substances and defective patterns.